New opportunity

Lead AI Security Engineer -APS

Societe Generale · Bangalore, India

About this role

The Lead AI Security Engineer - APS will design and develop AI-powered applications and agentic solutions within a hybrid work environment. Responsibilities include building RAG pipelines, developing agentic workflows using frameworks like LangChain and AutoGen, and implementing Model Context Protocol (MCP) servers. The role requires full-stack development expertise to integrate AI capabilities into production systems on AWS or Azure. Key Responsibilities: Design and develop AI applications using LLMs; build RAG pipelines including vector stores and embedding models; develop agentic solutions and orchestration patterns; evaluate LLM output quality and implement guardrails; perform full-stack development using Python/.NET and React; architect and deploy cloud-based solutions. Required Qualifications: 7–10 years of overall software development experience; 1–3 years of hands-on experience in the AI/ML domain; strong full-stack development skills (Python/.NET and React); experience with cloud platforms (AWS/Azure) and RAG pipeline development; familiarity with vector databases, agentic workflows, and REST API development. Preferred Qualifications: Familiarity with LLM observability tools (LangSmith, Arize, Helicone); knowledge of fine-tuning or RLHF; exposure to MCP server development; containerization experience (Docker/Kubernetes); and experience in the cyber domain.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

LLMsRAGLangChainLangGraphAutoGenCrewAISemantic KernelMCPPython.NETReactAWSAzureLlamaIndexChromaQdrantWeaviatepgvectorFastAPIFlaskASP.NETPrompt EngineeringLangSmithArizeHeliconeDockerKubernetes

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Architecting Agentic Workflows and RAG Pipelines
The role centers on designing and deploying AI-powered applications using frameworks like LangChain or AutoGen. You must demonstrate the ability to build robust RAG pipelines, manage vector stores, and implement complex orchestration patterns such as multi-agent systems and planning loops to ensure reliable tool-enabled LLM workflows.
Full-Stack Integration and Cloud Deployment
You are expected to bridge the gap between AI models and production systems. This requires proficiency in full-stack development (Python/.NET and React) and the ability to architect scalable solutions on AWS or Azure, ensuring that AI capabilities are securely and efficiently integrated into existing enterprise environments.
AI Quality Assurance and Security Guardrails
A critical priority is the evaluation of LLM output quality and the implementation of security guardrails. You must be prepared to manage prompt engineering, context window constraints, and observability to mitigate risks, especially given the emphasis on the cyber domain and secure AI application development.

Likely challenges to think through

Balancing Rapid AI Innovation with Enterprise Security
As an AI Security Engineer in a financial institution, you will face the challenge of implementing cutting-edge agentic solutions while adhering to strict security standards. Prepare by articulating how you balance the flexibility of LLM tool-use with the necessity of robust guardrails and data privacy.
Complexity of Multi-Agent Orchestration
Managing multi-agent systems involves significant technical overhead regarding memory management and tool-use reliability. Prepare by discussing your experience with state management in agentic workflows and how you troubleshoot non-deterministic behavior in complex LLM-based systems.

Practice your answers

Technical practice

Designing RAG Pipelines

What to coverExplain your approach to building a robust RAG pipeline, specifically addressing chunking strategies, embedding model selection, and vector database choice. Discuss how you ensure data retrieval relevance and minimize hallucinations in production environments.

Agentic Workflow Orchestration

What to coverDescribe your experience with agentic frameworks like LangChain or AutoGen. Explain how you manage multi-agent systems, planning loops, and memory management to ensure reliable task execution.

Model Context Protocol (MCP) Implementation

What to coverDiscuss your understanding of MCP servers and how they facilitate tool-enabled LLM workflows. Explain how you would expose internal APIs or data connectors to an agent runtime using MCP standards.

LLM Quality and Guardrails

What to coverHow do you evaluate the output quality of an LLM in a production system? Describe the guardrails you would implement to prevent prompt injection or data leakage, and suggest tools for monitoring these metrics.

Full-Stack AI Integration

What to coverAs a lead engineer, how do you architect the integration of AI services (e.g., Azure OpenAI or AWS Bedrock) into a full-stack application using Python/.NET and React? Focus on the communication patterns between the frontend and the backend AI services.

Cloud Deployment and Scalability

What to coverExplain your strategy for deploying AI applications on AWS or Azure. Discuss how you handle containerization with Docker and Kubernetes to ensure scalability and consistent performance for high-traffic AI services.

Prompt Engineering and Context Management

What to coverDescribe your methodology for optimizing prompts for complex tasks. Explain how you manage context window limitations when dealing with large datasets or long-running agentic conversations.

LLM Observability

What to coverWhich tools (e.g., LangSmith, Arize, Helicone) have you used to monitor LLM performance? Explain how you use these tools to debug agentic workflows and track cost or latency issues in production.

Behavioural practice

Handling Technical Trade-offs

What to coverDescribe a situation where you had to choose between different AI architectures or frameworks (e.g., choosing between LangChain and Semantic Kernel). Explain your decision-making process and the long-term impact on the project.

Cross-functional Collaboration

What to coverProvide an example of how you collaborated with data engineers or backend teams to integrate AI capabilities into an existing production system. How did you manage dependencies and ensure system stability?

Managing AI Security Risks

What to coverGiven the 'AI Security' focus of this role, describe a time you identified a potential security vulnerability in an AI application. What actions did you take to mitigate the risk and prevent future occurrences?

Mentorship and Technical Leadership

What to coverAs a lead engineer, describe a time you guided a team through the adoption of a new technology or framework. How did you ensure the team remained productive while learning the new stack?

Sharpen your knowledge

Topics to prepare

Agentic Workflow Orchestration
Review orchestration patterns including multi-agent systems, planning loops, and memory management. Exercise: Design a conceptual architecture for a multi-agent system where one agent retrieves data and another validates it against security guardrails, relevant to the role's focus on agentic solutions.
Retrieval-Augmented Generation (RAG) Architecture
Deepen understanding of RAG components including chunking strategies, embedding models, and vector store integration. Exercise: Map out a production-grade RAG pipeline from document ingestion to retrieval, focusing on how to maintain data security, which is critical for an AI Security Engineer.
Model Context Protocol (MCP)
Study the Model Context Protocol for standardizing tool-enabled LLM workflows. Exercise: Research the difference between SSE and stdio transport mechanisms in MCP servers to prepare for potential architectural discussions regarding tool exposure.

Skills to revise

Full-Stack AI Integration
Practice building REST APIs using Python (FastAPI/Flask) or .NET that interface with LLM backends and React frontends. Exercise: Create a simple React interface that consumes a Python-based API endpoint to trigger an LLM prompt, ensuring secure handling of API keys and user inputs.
LLM Guardrails and Evaluation
Gain hands-on experience with implementing guardrails to filter LLM outputs. Exercise: Implement a basic input/output validation layer in a Python script to detect and block prompt injection attempts or sensitive data leakage, directly supporting the security requirements of the role.
Cloud-Native Deployment
Familiarize yourself with deploying AI services on AWS or Azure. Exercise: Draft a deployment plan for a containerized RAG application using Docker, considering how to leverage cloud-native services like AWS Lambda or Azure AKS for scalable, secure execution.

Your preparation checklist

Architect a RAG Pipeline Project
Prepare a technical walkthrough of a RAG pipeline you have built. Focus on your choice of vector database (e.g., Chroma, Qdrant), embedding models, and chunking strategies. Be ready to explain how you handled data retrieval accuracy and context window management.
Demonstrate Agentic Workflow Expertise
Prepare a case study or code sample showcasing an agentic workflow using frameworks like LangChain, LangGraph, or AutoGen. Highlight your experience with orchestration patterns such as multi-agent systems, planning loops, and tool-use implementation.
Prepare Full-Stack Integration Evidence
Review your experience integrating AI capabilities into production systems. Be prepared to discuss how you connected Python or .NET backends with React frontends and deployed these services on AWS or Azure using containerization tools like Docker or Kubernetes.
Develop Model Context Protocol (MCP) Knowledge
Research and prepare to discuss your familiarity with MCP servers and clients. If you have experience with SSE or stdio transport for tool-enabled LLM workflows, document these specific implementations to demonstrate your technical depth.
Formulate Strategic Questions for the Interviewer
Prepare thoughtful questions regarding the team's approach to LLM observability (e.g., LangSmith, Helicone), how they implement security guardrails in production, and how they balance AI innovation with the cyber security requirements of a financial institution.

Availability has not yet been verified. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…